InfoResearchPeer-reviewed
SparseTG-GS: Test Case Generation via Sparse Perturbation Group
- Published
- Record updated
Summary
SparseTG-GS is a test case generation method for Deep Neural Networks in computer vision. It applies superpixel segmentation to an input image, iteratively adds perturbation groups until the target DNN is compromised, then prunes superfluous perturbations using an importance matrix from an improved AdvGAN-based GAN. The authors report the best attack success rate among compared methods, fewer and more interpretable perturbed superpixels, and efficiency on large images.
Related items
- MediumGHSA-25rm-9wvm-m38v: falcon-rust: Discrete-Gaussian sampler runs at 2^-33 precision, far below Falcon's requirement (reference KAT vectors no longer pass)Similar attack · GitHub Advisory Database
- InfoAdversarial Malware Detection Using Feature Contribution Difference Analysis with Kernel SHAPSimilar attack · OpenAlex (peer-reviewed AI security)
- InfoDensity Feedback Control for Transferability of Generative Sparse Adversarial AttacksSimilar attack · Arxiv (cs.RO + cs.CV security)
- InfoRECLAIM: Robot-Vision-Informed Adversarial Patch Defense via Embodied Attention FeedbackSimilar attack · Arxiv (cs.RO + cs.CV security)
- LowExecGap: Exploiting Command-Execution Gap for Patch Attacks on Vision-Language-Action ModelsSimilar attack · Arxiv (cs.RO + cs.CV security)